# Adaptive Sampling and Clipping for Private Worst-Case Group Optimization

arXiv: https://arxiv.org/abs/2602.10820 · HTML: https://mlsanity.com/paper/2602.10820/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 38 references that are in the map.
- First posted on arXiv 11 Feb 2026.

## Builds on (most cited first, 24 of 24)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [Deep Learning with Differential Privacy](https://mlsanity.com/paper/1607.00133/) (2016)
- [Group Normalization](https://mlsanity.com/paper/1803.08494/) (2018)
- [Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization](https://mlsanity.com/paper/1911.08731/) (2019)
- [Stochastic First- and Zeroth-order Methods for Nonconvex Stochastic Programming](https://mlsanity.com/paper/1309.5549/) (2013)
- [Renyi Differential Privacy](https://mlsanity.com/paper/1702.07476/) (2017)
- [Introduction to Multi-Armed Bandits](https://mlsanity.com/paper/1904.07272/) (2019)
- [Learning Models with Uniform Performance via Distributionally Robust Optimization](https://mlsanity.com/paper/1810.08750/) (2018)
- [Differential Privacy Has Disparate Impact on Model Accuracy](https://mlsanity.com/paper/1905.12101/) (2019)
- [Subsampled R'enyi Differential Privacy and Analytical Moments Accountant](https://mlsanity.com/paper/1808.00087/) (2018)
- [Large-Scale Methods for Distributionally Robust Optimization](https://mlsanity.com/paper/2010.05893/) (2020)
- [How to DP-fy ML: A Practical Guide to Machine Learning with Differential Privacy](https://mlsanity.com/paper/2303.00654/) (2023)
- [Revisiting Gradient Clipping: Stochastic bias and tight convergence guarantees](https://mlsanity.com/paper/2305.01588/) (2023)
- [Differential Privacy and Machine Learning: a Survey and Review](https://mlsanity.com/paper/1412.7584/) (2014)
- [Differentially Private and Fair Deep Learning: A Lagrangian Dual Approach](https://mlsanity.com/paper/2009.12562/) (2020)
- [Turning the Tables: Biased, Imbalanced, Dynamic Tabular Datasets for ML Evaluation](https://mlsanity.com/paper/2211.13358/) (2022)
- [Disparate Impact in Differential Privacy from Gradient Misalignment](https://mlsanity.com/paper/2206.07737/) (2022)
- [Privacy amplification by random allocation](https://mlsanity.com/paper/2502.08202/) (2025)
- [Recent Advances of Differential Privacy in Centralized Deep Learning: A Systematic Survey](https://mlsanity.com/paper/2309.16398/) (2023)
- [Subsampling is not Magic: Why Large Batch Sizes Work for Differentially Private Stochastic Optimisation](https://mlsanity.com/paper/2402.03990/) (2024)
- [Differentially Private Worst-group Risk Minimization](https://mlsanity.com/paper/2402.19437/) (2024)
- [Communication-efficient Decentralized Local SGD over Undirected Networks](https://mlsanity.com/paper/2011.03255/) (2020)

## Abstract (arXiv)

A central requirement for the acceptance of machine learning methods for human-centric tasks is that they should be fair, in the sense that they should work comparably well for individuals from different societal groups. A second, equally important, requirement is that they should respect the privacy of user data. While techniques exist to address each aspect in isolation, such as worst-case group optimization for the former and differentially private SGD for the latter, these are often at odds with with each other, and no practical method currently exists to enforce both requirements simultaneously. In this work, we overcome this problem and propose an algorithm for optimizing the worst-case group accuracy in a differentially private way. Our main contribution is ASC (Adaptively Sampled and Clipped Worst-case Group Optimization), which adaptively controls both the sampling rate and the clipping threshold of each group's gradient contributions. Thereby, it is able to reweight the training objective in favor of harder-to-learn groups, while keeping the noise required to enforce privacy low enough to preserve model utility. Our experiments show that ASC achieves substantially higher worst-case group accuracy than prior work, without sacrificing overall average accuracy.